PL242373B1 - Method of generating music content - Google Patents

Abstract

Music content generation starts with input music content (1). The next step in initiating the generation of music content is to decide whether to analyze the input music content manually or automatically. Regardless of the form of analysis (3) of the input music content (1), on the basis of business rules and composition rules (4), models of music generation are developed (5) and a melody generator is created (10). At the same time (2), sounds (6) are prepared and can be stored in the sound repository (8). Then the models in the form of source code are transferred to the generator (14). First, the generator is set (12) with specific parameters using a control device (26) compatible with MIDI standards and supplemented with the features of the song read from the user's preferences database (13). Then, this content is sent to automatic generation (15) based on artificial intelligence algorithms and a digital score of the song with the desired characteristics is generated (15). The soundtracks of the individual instruments are rendered (16) and the rendered tracks are mixed into the final music recording (20). Then, the work (27) and its recording (20) are verified by the critic module (23) based on algorithms based on neural networks. After positive evaluation, the work is exported (23) to a dedicated platform (24).

Description

Description of the invention

The subject of the invention is a method of generating musical content based on algorithms of successive processes on input data sets.

In this patent description, the term algorithms should be understood as successive operations on data in the form of a set of musical content created on the basis of composition rules, created on the basis of user preferences as well as business and legal rules standardizing the organization of the process of generating musical content and their recordings.

The concept of business rules should be understood as a statement that defines an aspect of business aimed at controlling or influencing business behavior by informing that there are specific requirements regarding conduct, actions and practices or procedures implemented as part of a given user activity. For the purpose of this description of the solution, business rules were combined with the rules of music composition.

The concept of business rules with the rules of music composition should be understood as a set of rules for musical compositions, taking into account the sales estimation of musical forms and the standards of musical composition.

A repository of atomic sounds is a collection of recordings of single notes.

In the process of generating musical content, the algorithms carry out a number of processes aimed at the first stages at creating a generator that generates a digital notation of the model, which is given the parameters and features of the song in the generator in accordance with the user's preferences. In this area of the generator's operation, musical content is created at the level of a technical piece.

At the technical level, the term musical content should be understood as the developed model, which is the result of a series of processes aimed at creating a generator. Creating the generator is divided into two stages, where in the first sub-stage the content is created at the abstract level as an element of form, in the second sub-stage the abstract content is converted to a digital notation of each line of a given instrument separately, containing form, harmony and layers, shaping the melody by task the generator with the parameters and features of the work in accordance with the user's preferences.

In the second stage of the process of generating musical content, the notation is processed in such a way that it turns into a recording in a specific sound form using sampling synthesis and samples as recordings of individual sounds. The concept of sampling synthesis is understood as the process of creating musical content using a fragment of a previously made music recording, called a sample, as an element of a newly created work.

At the rendering stage, the musical content in the form of a model is converted by the sequencer and sampler using samples into the form of recording individual instruments separately. The content created in this way is subjected to the next stage of the process of mixing and refining the recording. The final recording is obtained, which is verified by the critic module based on artificial intelligence algorithms of neural networks. At this stage, an artistic work is obtained with compositional values in accordance with business standards and compositional rules. The track is then exported to the platform's distribution module.

A number of solutions for generating musical works are known. A known solution is presented in the patent document KR 20190100543. It discloses an electronic device comprising: a display, a processor electrically connected to the display, a memory electrically connected to the processor, whereby when processing is performed, the processor checks certain input parameters and output audio tracks and, based on an artificial intelligence algorithm, it identifies sound parameters. The electronic device stores instructions for checking the appropriate composition information and displaying assistance information for correcting the input audio track on the display based on the confirmed composition information. The defined parameters contain specific information, in particular about the musical style, information about the musical instrument, information about the rhythm, information about the time signature and information about the music genre.

Another known solution is presented in the patent document CN 110211556. It presents a method and device for processing a music file, a terminal and a storage medium. The processing method includes the following steps: the first human voice data is collected to input the target sounds, the reverberation parameters obtained by the target human voice data corresponding to the target music are obtained. The first human voice data is processed based on reverberation parameters and the second human voice data is obtained. Secondly, the human voice data is processed and the accompaniment data corresponding to the target music is synthesized and the target music file is obtained.

A further known solution is presented in the patent description KR 20190105254. A solution is described herein which aims to provide a fully digital audio processing device that directly receives a digital audio source signal. The processing is based on the file containing the digital audio source, the audio signal processing is done in a fully digital way in conjunction with the input signal of the digital audio source. The audio signal is then sent to the loudspeaker. After directly inputting a digital signal from a digital audio source file, all signal processing can still be done digitally, and the audio signal can be adaptively processed based on artificial intelligence.

According to the invention, a method of generating musical content based on a series of successive processes, the operation and course of which is based on the work of artificial intelligence algorithms. The process of generating music content is carried out using a control controller corresponding to the MIDI standards. Business rules have been created to automatically build music tracks based on user preferences. Automatic generation of music content is possible through such solutions as the user's preferences database, repository resources, business rules, models of generation of music pieces of a given type and a melody generator in which parameters and features for models for the form and line of instruments are provided. Models are created at the technical level, which are further processed on the basis of modification algorithms of musical input files to generate the final recording, and after its verification, a piece containing the intended compositional and artistic load.

The method of generating music content according to the invention is characterized in that the input sound samples are processed on the basis of modification algorithms of the music input files, in particular regarding features such as tempo, mood of the song, musical genre, duration and content modulation range. The result is a composition with the intended artistic expression. The generation process in the first stage consists in building musical content at the technical level in the form of models. Technical content is obtained as a result of a series of processes focused on creating a generator. Carrying out a series of processes consists in the fact that after entering the input content, the input music content is analyzed for the occurrence of patterns. Then, the patterns are stored in the database of business rules and rules of music composition, on the basis of which models of generation of music pieces of a given type are developed. This is how a melody generator is created, in which a digital notation of the lines of a given instrument is generated. At the same time, a database of atomic sounds is prepared and then transferred to the generator, where the parameters of the generator are set using a control device compatible with MIDI standards. The resulting models are automatically generated with digital music notation and parts for individual instruments are created, which are then rendered to the music tracks for each instrument. You get a recording on an artistic level. The recording is then refined and mixed. The final version of the recording is obtained, after which the work and its recording are verified by the critic module. After verification, the recording is exported to the distribution module of a dedicated platform.

In the preferred version of the solution, the final music recording is created using artificial intelligence algorithms at the stage of analyzing for existing patterns, developing song generation models, creating a melody generator, preparing sounds.

In another preferred version of the solution, sound samples are created as content is written to the repository.

In another advantageous version of the solution, the developed models are transferred for reading and a digital score of the work with the required features is automatically generated.

In another preferred embodiment, instrument soundtracks are rendered using repository resources.

In another advantageous version of the solution, the song and its recording are verified using artificial intelligence and the process of generating musical content is repeated from the beginning.

By using ready-made patterns of patterns and samples, the user without special instrumental and hardware resources and without substantive knowledge at the level of a programmer or sound engineer, using a control controller to determine the characteristics of the sound content of the sound content, will be able to create full-scale musical content with artistic values, refined according to individually defined compositional preferences.

Artificial intelligence algorithms are used in the process of creating music content, which results in the work of a team of specialists responsible for generating this type of music content in a traditional way. The operation of the generator is supported and controlled by a control device in the MIDI standard. Fully digital generation of music content using a control device gives the user the ability to specify the guidelines for the generator in the form of setting basic parameters, in particular for the genre, tempo, mood, duration and content modulation parameters that give individual content. The user's work is additionally supported by a sound repository containing sounds in the form of single notes. Music tracks for individual instruments are rendered to a form that is then mixed and refined to the level according to the intended artistic composition. An algorithm based on the work of extensive relay networks checks the work and its recording for compliance with compositional assumptions, in particular compliance with preferences and business standards applicable to composition. Music content can be generated without limits. The process of creating a generator is one-time. The generated music content can be distributed.

The subject of the invention is presented in an embodiment in the attached drawing which illustrates an exemplary block diagram of generating music content.

The block diagram shows the course of subsequent steps for the implementation of the subject of the invention, and also indicates the sets and databases used when generating new music content with the method according to the invention. In this description, the terms musical content, work and work and its recording are used to denote the result of the procedure according to the invention. An element necessary to implement the method according to the invention is a control device compatible with MIDI standards.

In the block diagram shown in the figure, each "+" symbol should be understood as a conjunction of a series of processes following each other at one time.

Arrows marked along the length with a dashed line should be understood as an indication of the course of activities earlier in time to the course of activities of the arrows marked along the length with solid lines.

Each first arrow leading to the base tile 25 should be understood as a "writes to", each arrow drawn from the base tile 25 should be understood as "reads to".

Existing works are understood as existing sound compositions or sound samples.

The term "work and its recording" at the stage of export to the distribution module of platform 23 means that not only the recording itself is verified, but also, among others, part of the information provided by the user on the parameters and features of the work 12, including the concept of the work, e.g. its genre.

The concept of content at the technical level 26 should be understood as a MIDI file and additional data transferred to the generator in the form of technical algorithms and source code.

The term sequencer 28a is understood as an electronic device or a computer program that stores a sequence of sounds, but a sequence of synthesizer control instructions along with parameters and enables its repeated playback.

The term samplers 28b is understood as an electronic musical instrument or a computer program that enables digital recording of any sound and then using it as a traditional musical sound.

The area of operation of the sampler and sequencer is understood as the combined operation of the modules: 28a and 28b on MIDI files and data on musical content, constituting instructions for the rendering process 16.

The concept of verification by the Al 19 critic module should be understood as verifying the recording and its work through a module based on the operation of artificial intelligence algorithms based on extensive neural networks. These are neural network learning algorithms that are primarily capable of generalizing the observed data. Network learning is understood as forcing the network to respond in a specific way to given input signals.

As shown in Fig. 1, the process of generating the musical content starts with the creation of the generator, where at the beginning the input musical content 1 from the existing works is analyzed. The two-way process consists in the fact that existing works are analyzed at the same time in terms of the occurrence of patterns of 3 existing works and models of generation of musical works 5 are developed, a melody generator 10 is created and sounds 6 are prepared. of the source code are fed into a generator 14 which is given parameters and features. Setting the characteristics and parameters of the generator is preferably done by means of a control device 26 compatible with MIDI standards.

The content from the generator as models at the technical level 27 is transferred to the generation where automatic generation of the digital score of the work with the desired features 15 takes place, based on artificial intelligence algorithms, and then the sound parts for individual instruments are obtained 17. The resulting sound parts 17 are transferred as information that is analyzed in the area of operation of the sequencer and sampler 28 and renders 16 this content separately for each instrument, in this way, using the sequencer and sampler with the use of samples, the digital notation of each line of a given instrument is changed into a sound form and a form of recording is created individual instruments separately. Then the 18th recording is refined and mixed. In this way, a final music recording 20 is obtained which is submitted for verification. Work 27 and its recording 20 are verified using the critic module 19 based on specialized neural network algorithms. The final musical content is exported 23 and sent for distribution 24. In the case of a negative verification of the work and its recording in the criticism module 19, the process stops at this stage and the automatic generator 16 generates new content based on the parameters and features set for it, preferably user preference database 13.

In this embodiment of the solution according to the invention, in the sound preparation process step 6, the prepared musical content is stored in the sound repository 8.

In this embodiment, in the solution according to the invention, the generator is asked the parameters and features of the song from the user's preference database 13.

In another embodiment, generation models of musical pieces of a given type are developed and stored in a database 11 of developed models from which these models are read at the stage of automatic generation of a digital score of a piece with the desired characteristics 15.

Claims (6)

Patent claims 1. The method of generating music content according to the invention, where the input sound samples are processed on the basis of modification algorithms of the music input files, concerning in particular features such as tempo, mood of the song, music genre, duration, and the range of content modulation is selected, where as a result it is obtained a composition with an intended artistic expression, characterized in that musical content is built at the technical (27) and artistic (20) level, where at the level of content creation at the technical level (20) the input musical content (1) is analyzed in terms of the occurrence of patterns (3), patterns are recorded in the database of business rules and rules of music composition (4) on the basis of which generation models of music pieces of a given type are developed (5), then a melody generator is created, in which a digital notation of a line of a given instrument is generated ( 10), while at the same time a database of atomic sounds (6) and then the content of the muse are being prepared The data is transferred to the generator (14) in which, using a control device (26) compliant with MIDI standards, parameters (12) are assigned to the generator and the digital score of the work is automatically generated (15) and parts for individual instruments are created (17) which are then rendered (16) to the music tracks for each instrument, and then the individual tracks are mixed into a recording (18) and the final version of the recording is obtained (20), after which the song (27) and its recording (20) are subjected to verified by the Al (19) critic module. 2. A method for generating music content according to claim 1. 1, characterized in that the final music recording (20) is created with the use of artificial intelligence algorithms at the stage of analyzing for the occurrence of existing patterns (3), models of song generation are developed (5) and a generator (6) is created for the preparation of sounds ( 10). 3. A method for generating music content according to claim 1. The method according to any of claims 1 and 2, characterized in that sound samples (6) are created and the content is saved in the repository (8) in parallel. 4. A method for generating music content according to claim 1. The process according to claim 1, characterized in that the developed (5) models (11) are sent for reading and a digital score of the work with the required features (15) is automatically generated. 5. A method for generating music content according to claim 1. The instrument according to claim 1, characterized in that the instrument soundtracks (17) are rendered (16) using repository resources (8). 6. A method for generating music content according to claim 1. 1, characterized in that the work and its recording are verified (19) using artificial intelligence algorithms and the process of generating musical content from the beginning is repeated.